Intelligent Attachment Control Method for Flexible Lander

Abstract

The intelligent attachment control method for a flexible lander disclosed by the present invention belongs to the technical field of deep space exploration. The implementation method of the present invention is as follows: the overall attitude of the flexible lander is characterized by the height difference between thrust nodes, and the attitude control problem of the flexible lander is transformed into the control problem of the height difference between nodes; an analytical nominal-intelligent compensation node height difference control law is designed, the analytical nominal control term is determined by feedback of the node height difference and its change rate, and the intelligent compensation term is designed by neural network learning of the flexible internal force difference, so as to obtain an analytical nominal-intelligent compensation node height difference control law with the node thrust difference as the control quantity; by designing a node thrust distribution law, the control quantity of the node thrust difference is reasonably distributed to each node thruster. The flexible lander applying the present invention can coordinate the control capabilities of each thruster, realize overall attitude maneuvering and attitude maintenance, and the control process has good transient performance and steady-state performance.

Description

Technical Field

[0001] The present invention relates to a method for intelligent control of a lander, and particularly to a method for intelligent attachment control of a flexible lander, belonging to the technical field of deep space exploration. Background Art

[0002] Due to the characteristics of small celestial bodies such as weak gravity, complex environmental disturbances, and irregular surface topographies, traditional rigid landers with a cube configuration have a relatively high risk of overturning and escaping during landing. A new type of lander with a disk-shaped configuration and flexible materials can increase the contact area with the surface of small celestial bodies during landing, and use internal damping to consume the residual kinetic energy during landing, thereby reducing the risk of overturning and rebounding and improving the landing success rate. The overall configuration of the flexible lander is disk-shaped and consists of multiple rigid mass aggregation areas wrapped and connected by flexible materials. The thrusters and sensors of the flexible lander are all installed in the rigid mass aggregation areas. During the landing process, in order to maintain or adjust the pointing of the sensors and thrusters, it is necessary to maintain or adjust the attitude of the flexible lander. Different from the attitude control of traditional rigid landers, the attitude control of flexible landers faces the following difficulties: on the one hand, the flexible lander is an infinite-dimensional distributed parameter system, and the overall attitude is difficult to characterize; on the other hand, the flexible lander is controlled by multiple pairs of thrusters in cooperation, and the flexible connection between the thrusters causes the thrust action to be coupled, making it difficult to allocate attitude control commands. For flexible landers, it is very necessary to study reasonable attitude characterization methods, design reasonable control laws and distribution laws, and coordinate the thrusters to complete the control of the lander's attitude maintenance and attitude maneuvering to achieve safe and stable landing of the flexible lander on the surface of small celestial bodies. Summary of the Invention

[0003] Aiming at the special configuration of the flexible lander and the attitude control requirements during the landing process, the main purpose of the present invention is to provide a method for intelligent attachment control of a flexible lander, which characterizes the overall attitude of the flexible lander through the height difference between thrust nodes, and transforms the attitude control problem of the flexible lander into a control problem of the height difference between nodes; designs an analytical nominal-intelligent compensation node height difference control law, designs an analytical nominal control term by feedback of the node height difference and its change rate, determines the intelligent compensation term through neural network learning, and obtains an analytical nominal-intelligent compensation node height difference control law with the node thrust difference as the control quantity; realizes the reasonable allocation of the node thrust difference control quantity to each node thruster by designing a node thrust distribution law. Applying the method for intelligent attachment control of the flexible lander, the flexible lander can coordinate the control capabilities of each thruster, complete overall attitude maneuvering and attitude maintenance, and the control process has good transient performance and steady-state performance.

[0004] The object of the present invention is achieved by the following technical solutions.

[0005] The intelligent attachment control method for a flexible lander disclosed in the present invention includes the following steps:

[0006] Step 1. Aiming at the problems that the flexible lander system is high-dimensional and complex, and the overall attitude is difficult to characterize and control, the overall attitude of the flexible lander is described by the height difference between the thrust nodes, the conversion relationship between the pitch and roll attitudes of the flexible lander and the node height difference is deduced, and the attitude control problem of the flexible lander is transformed into the control problem of the node height difference.

[0007] The specific implementation method of Step 1 is as follows:

[0008] There are mass concentration areas of thrusters and sensors on the flexible lander, which are called nodes. The three-node flexible lander is in a standard disk shape in the nominal state without deformation. The three nodes are evenly distributed on the surface of the detector, forming an equilateral triangle. The three nodes are numbered 1, 2, and 3 in counterclockwise order, called Node 1, Node 2, and Node 3.

[0009] On the premise of retaining the flexible characteristics and the planar configuration characteristics, a simplified model of the flexible lander oriented to the nodes is established.

[0010] Define the landing point coordinate system o L -x L y L z L , with the landing point o L as the origin, the normal direction perpendicular to the surface of the small celestial body at the landing point is the z L axis, the east direction of the small celestial body surface is the x L axis, and the north direction of the surface is the y L axis. In the landing point coordinate system, the node position vector r i of the flexible lander is

[0011]

[0012] where i is the thrust node number (i = 1, 2, 3), x i is the component of the line connecting node i and the landing point on the x L axis, y i is the component of the line connecting node i and the landing point on the y L axis, and z i is the component of the line connecting node i and the landing point on the z L axis, that is, the height of node i on the surface of the small celestial body.

[0013] The mass of the flexible lander is concentrated at each node, and the center of mass of the flexible lander is the centroid of the node triangle

[0014]

[0015] where r cis the position vector of the center of mass of the flexible lander.

[0016] For a flexible lander with a circular disk configuration, only consider the attitude tilt along the radial direction of the circular disk. Taking the line connecting the node centers of mass as the axis, define the pitch attitude and roll attitude of the flexible lander. Using the vector r from the center of mass of the flexible lander to node 1 c1 as the axis, the pitch attitude of the flexible lander is represented by the angle between r c1 and o L x L y L plane, and the expression for the pitch angle θ is

[0017]

[0018] The roll attitude of the flexible lander is represented by the angle between the vector r from node 3 to node 2 32 and o L x L y L plane, and the expression for the roll angle ψ is

[0019]

[0020] During the landing process, the change in the distance between nodes is much smaller than the original length of the node distance. When calculating the pitch and roll attitude angles, ignore the change in the distance between nodes, and the approximate expressions for the pitch angle and roll angle are

[0021]

[0022] where L0 is the original length of the node distance under the nominal state.

[0023] Define the height difference variable H between the nodes of the flexible lander

[0024]

[0025] where h 12 is the height difference between node 1 and node 2, and h 13 is the height difference between node 1 and node 3.

[0026] The pitch and roll attitudes of the disk-shaped flexible lander are characterized by the height difference between the nodes, and the conversion relationship between the height difference variable and the pitch angle and roll angle is

[0027]

[0028] Thus, the overall attitude control problem of the flexible lander is transformed into a control problem of the node height difference.

[0029] Step 2: Establish a node-oriented flexible lander dynamics model. With the node thrust difference as the control variable, design an analytical nominal-intelligent compensation node height difference control law. Based on a PD controller, implement feedback control of the node height difference and the change rate of the node height difference, and determine the analytical nominal control term in the node height difference control law.

[0030] The specific implementation method of Step 2 is as follows:

[0031] In the landing point coordinate system, the landing dynamics model of the node-oriented flexible lander is

[0032]

[0033] where v i =[v xi , v yi , v zi T is the velocity of node i, u i =[u xi , u yi , u zi T is the thrust of node i, m is the mass of each node, f i =[f xi , f yi , f zi T is the flexible internal force acting on node i, and a ai =[a axi , a ayi , a azi T is the acceleration of node i caused by the small celestial body environmental force.

[0034] Since the size of the flexible lander is much smaller than the size of the small celestial body, it is considered that the external environmental forces acting on each node are equal. From the landing dynamics, the acceleration difference between the node height directions is only related to the thrust difference and the flexible internal force difference in the node height direction, that is

[0035]

[0036] where U 12 is the difference in the thrusts received by node 1 and node 2 in the height direction, U 13 is the difference in the thrusts received by node 1 and node 3 in the height direction, f 12 is the difference in the flexible internal forces received by node 1 and node 2 in the height direction, and f 13 is the difference in the flexible internal forces received by node 1 and node 3 in the height direction, that is

[0037]

[0038] ​​​​Based on the dynamic relationship between the acceleration difference in the node height direction and the thrust difference and flexible internal force difference in the node height direction, with the node thrust difference as the control variable, an analytical nominal-flexible compensation node height difference control law is designed, and the expression is

[0039]

[0040] where, is the analytical nominal control term of the node height difference, is the compensation term for f 12 , is the compensation term for f 13 .

[0041] Based on the PD controller, a feedback control for the node height difference and the change rate of the node height difference is designed, and the analytical nominal control term in the node height difference control law is established

[0042]

[0043] where, k p > 0 is the gain of the height difference feedback term, k d > 0 is the gain of the height difference change rate feedback term, H e is the desired height difference, is the desired height difference change rate.

[0044] Step 3: Based on neural network learning, an intelligent mapping function between the node state variables and the node flexible internal force difference is established as the intelligent compensation term in the node height difference control law. Through the weighted switching coefficient of the intelligent compensation term, it is avoided that the flexible internal force is over-compensated after the control of the flexible lander enters the steady state. Combining with the analytical nominal control term established in Step 2, an analytical nominal-intelligent compensation node height difference control law is obtained.

[0045] The specific implementation method of Step 3 is as follows:

[0046] During the simulated landing process of the flexible lander, state variable sampling is carried out to generate the data set required for training the neural network. Periodic random thrusts in the height direction are applied to the three thrust nodes of the flexible lander. The thrust amplitude of each node is limited, and the value range is u zi ∈[-u max , u max , where u max is the thrust constraint upper limit value. The action period of each group of random thrusts is ΔT, and then the next group of random thrusts is switched until the pitch angle or roll angle of the flexible lander reaches the constraint upper limit values θ max , ψ max , or the flight time reaches the constraint upper limit value T max. During the landing process, the node position, node attitude, and internal forces acting on the nodes of the flexible lander are sampled at intervals of time δt, and the data is recorded and saved. The above landing sampling process is repeated N times to generate a mapping dataset of node state variables and node internal forces.

[0047] During the landing process of the flexible lander, the internal forces acting on each node are related to the height difference between nodes, the relative distance between nodes, and the attitude of the nodes themselves. Taking the node state variables and node internal force differences in the mapping dataset as input and output variables, based on neural network learning, an intelligent mapping function for node internal forces is established. The expression of the mapping function is:

[0048]

[0049] where, f in (·) is the intelligent mapping function, the node distance deformation variable δL = [L 12 -L0, L 13 -L0, L 23 -L0] T , L 12 is the distance between node 1 and node 2, L 13 is the distance between node 1 and node 3, L 23 is the distance between node 2 and node 3. The node angular acceleration variable is the component of the attitude angular acceleration of node i in the x L axis and y L axis of the landing point coordinate system.

[0050] When the flexible lander performs an attitude tilt maneuver, the relative distance between nodes and the node attitude change rapidly, and the nodes are subjected to strong flexible internal forces. At this time, internal force compensation is considered in the control design to make the node height difference control process of the flexible detector have better transient performance. When the flexible lander tends to reach a steady state, the overall lander shows strong rigidity, and the flexible internal forces will be overcompensated, resulting in a large steady-state error. Therefore, through the weighted switching coefficient of the internal force intelligent compensation term, the compensation of the flexible internal forces in the attitude control process is made more reasonable.

[0051] For the control process of the attitude tilt maneuver of the flexible lander, the expression of the weighted switching coefficient λ is

[0052]

[0053] where, e max represents the criterion value for the height difference control to enter the steady state. After the attitude maneuver of the flexible lander enters the steady state, the weighted coefficient value is switched to zero, and the flexible internal forces between nodes are no longer compensated.

[0054] With the node thrust difference as the control variable, the analytical nominal-intelligent compensation node height difference control law is finally written as

[0055]

[0056] Step 4: Design the node thrust distribution law to distribute the node thrust difference control variable in the analytical nominal-intelligent compensation node height difference control law established in Step 3 to each node thruster, so that each node thruster with limited thrust amplitude can make full use of its own control ability to achieve intelligent cooperative control of the overall attitude of the flexible lander.

[0057] The specific implementation method of Step 4 is as follows:

[0058] Design the node thrust distribution law to distribute the node thrust difference control variable to the thrusters of each node. To make full use of the control ability of each node thruster, the maximum value of the node thrust and the minimum value of the node thrust should be opposite to each other, that is, u i =-u k when u i ≤u j ≤u k , i, j, k = 1, 2, 3. The expression of the node thrust distribution law is

[0059]

[0060] According to formula (16), distribute the node thrust difference control variable to each node thruster, so that each node thruster with limited thrust amplitude can make full use of its own control ability to achieve intelligent cooperative control of the overall attitude of the flexible lander.

[0061] Beneficial effects:

[0062] 1. For the problem that the overall attitude of the flexible lander is difficult to characterize and control due to its complex high dimension and the coupling of thrust actions, the intelligent attachment control method of the flexible lander disclosed in the present invention establishes an attitude characterization model of the flexible lander for the node height difference, and transforms the overall attitude control of the flexible lander into the control of the node height difference, making the attitude maneuver control problem of the flexible lander simple and solvable.

[0063] 2. The intelligent attachment control method of the flexible lander disclosed in the present invention realizes the control of the overall attitude of the flexible lander by designing the analytical nominal-intelligent compensation node height difference control law. By designing the weighted switching coefficient of the intelligent compensation term, it is avoided that the flexible internal force is over-compensated after the control of the flexible lander enters the steady state, making the attitude maneuver control process of the flexible lander have good transient performance and steady state performance.

[0064] 3. The intelligent attachment control method of the flexible lander disclosed in the present invention realizes the reasonable distribution of the node thrust difference control quantity in the node height difference control law to each node thruster by designing the node thrust distribution law. Under the condition that the node thrust amplitude is limited, the full utilization of the thrust capacity of each node is realized, and the overall attitude maneuver control of the flexible lander is completed by coordinating each thruster. Description of the Drawings

[0065] Figure 1 It is a schematic flow chart of the intelligent attachment control method of the flexible lander disclosed in the present invention.

[0066] Figure 2 It is a schematic diagram of the node height difference.

[0067] Figure 3 It is the node height difference curve of the pitch maneuver of the flexible lander.

[0068] Figure 4 It is the pitch angle curve of the pitch maneuver of the flexible lander.

[0069] Figure 5 It is the three-node thrust curve of the pitch maneuver of the flexible lander. Detailed Implementation Modes

[0070] In order to better illustrate the purpose and advantages of the present invention, the content of the invention will be further described below in conjunction with the drawings and examples.

[0071] In order to verify the feasibility of the method, taking the landing mission of the flexible lander on the small celestial body Eros433 as an example, the simulation of the intelligent attachment control method of the flexible lander is carried out. The mass of the flexible lander node m = 333 kg, the original length of the distance between nodes L0 = 1.0392 m, the gain k of the analytical feedback control term p = 0.15, k d = 0.5, the upper limit of the node thrust u max = 10 N, and the steady-state criterion value of the height difference e max = 0.01 m.

[0072] As Figure 1 shown, the specific implementation steps of the intelligent attachment control method of the flexible lander disclosed in this embodiment are as follows:

[0073] Step 1: Aiming at the problems that the flexible lander system is high-dimensional and complex, and the overall attitude is difficult to characterize and control, a scheme is proposed to describe the overall attitude of the flexible lander by the height difference between the thrust nodes, the conversion relationship between the pitch and roll attitudes of the flexible lander and the node height difference is deduced, and the attitude control problem of the flexible lander is transformed into the control problem of the node height difference.

[0074] The specific implementation method of Step 1 is:

[0075] The mass concentration areas on the flexible lander equipped with thrusters and sensors are called nodes. The three-node flexible lander is in the standard disk shape in the nominal state without deformation. The three nodes are evenly distributed on the surface of the detector, forming an equilateral triangle. The three nodes are numbered 1, 2, and 3 in the counterclockwise order, called Node 1, Node 2, and Node 3.

[0076] On the premise of retaining the flexible characteristics and the planar configuration characteristics, a simplified model of the flexible lander oriented to nodes is established.

[0077] Define the landing point coordinate system o L -x L y L z L , with the landing point o L as the origin, the normal direction perpendicular to the surface of the small celestial body at the landing point is the z L axis, the eastward direction on the surface of the small celestial body is the x L axis, and the northward direction on the surface is the y L axis. In the landing point coordinate system, the node position vector r i of the flexible lander is

[0078]

[0079] where i is the thruster node number (i = 1, 2, 3), x i is the component of the line connecting node i and the landing point on the x L axis, y i is the component of the line connecting node i and the landing point on the y L axis, and z i is the component of the line connecting node i and the landing point on the z L axis, that is, the height of node i on the surface of the small celestial body.

[0080] The mass of the flexible lander is concentrated at each node. Ignoring the mass of the flexible material and the mass difference at each node, it is considered that the center of mass of the flexible lander is the centroid of the node triangle

[0081]

[0082] where r c is the position vector of the center of mass of the flexible lander.

[0083] For the flexible lander with a circular disk planar configuration, only consider the attitude tilt along the radial direction of the circular disk. Taking the line connecting the node centroids as the axis, define the pitch attitude and roll attitude of the flexible lander. Taking the vector r c1 from the center of mass of the flexible lander to Node 1 as the axis, the pitch attitude of the flexible lander is represented by the angle between r c1 and o L x L yL represented by the included angle of the plane, and the expression of the pitch angle θ is

[0084]

[0085] The roll attitude of the flexible lander is represented by the vector r pointing from node 3 to node 2 32 and o L x L y L represented by the included angle of the plane, and the expression of the roll angle ψ is

[0086]

[0087] During the landing process, the change in the distance between nodes is much smaller than the original length of the node distance. When calculating the pitch and roll attitude angles, the change in the distance between nodes is ignored, and the approximate expressions for the pitch angle and roll angle are

[0088]

[0089] where L0 is the original length of the node distance under the nominal state.

[0090] Define the height difference variable H between the nodes of the flexible lander

[0091]

[0092] where h 12 is the height difference between node 1 and node 2, and h 13 is the height difference between node 1 and node 3, as Figure 2 shown.

[0093] The pitch and roll attitudes of the disk-shaped flexible lander are characterized by the height difference between the nodes, and the conversion relationship between the height difference variable and the pitch angle and roll angle is

[0094]

[0095] Thus, the overall attitude control problem of the flexible lander is transformed into a control problem of the node height difference.

[0096] Step 2: Establish a node-oriented dynamic model of the flexible lander, use the node thrust difference as the control quantity, and design an analytical nominal-intelligent compensation node height difference control law. Based on the PD controller, design the feedback control of the node height difference and the change rate of the node height difference as the analytical nominal control term in the node height difference control law.

[0097] The specific implementation method of Step 2 is as follows:

[0098] In the landing point coordinate system, the landing dynamics model of the node-oriented flexible lander is

[0099]

[0100] Among them, v i =[v xi , v yi , v zi T is the velocity of node i, u i =[u xi , u yi , u zi T is the thrust of node i, m is the mass of each node, f i =[f xi , f yi , f zi T is the flexible internal force acting on node i, a ai =[a axi , a ayi , a azi T is the acceleration generated by node i under the small celestial body environmental force (including small celestial body gravity and small celestial body spin inertia force, etc.).

[0101] Since the size of the flexible lander is much smaller than the size of the small celestial body, it is considered that the external environmental forces acting on each node are equal. From the landing dynamics, the acceleration difference between the node heights is only related to the thrust difference and the flexible internal force difference in the node height direction, that is

[0102]

[0103] Among them, U 12 is the thrust difference between node 1 and node 2 in the height direction, U 13 is the thrust difference between node 1 and node 3 in the height direction, f 12 is the flexible internal force difference between node 1 and node 2 in the height direction, f 13 is the flexible internal force difference between node 1 and node 3 in the height direction, that is

[0104]

[0105] Based on the dynamic relationship between the acceleration difference between the node heights and the thrust difference and the flexible internal force difference of the nodes, with the node thrust difference as the control variable, an analytical nominal-flexible compensation node height difference control law is designed, and the expression is

[0106]

[0107] Among them, is the analytical nominal control term of the node height difference, is for f 12 ​​​​The compensation term, is f 13 The compensation term.

[0108] Based on the PD controller, design the feedback control for the node height difference and the change rate of the node height difference, and establish the analytical nominal control term in the node height difference control law

[0109]

[0110] where k p = 0.15 is the gain of the height difference feedback term, k d = 0.5 is the gain of the change rate of the height difference feedback term, H e is the desired height difference, is the desired change rate of the height difference.

[0111] Step 3: Based on neural network learning, establish the intelligent mapping function between the node state quantity and the node flexible internal force difference, as the intelligent compensation term in the node height difference control law. Design the weighted switching coefficient of the intelligent compensation term to avoid over-compensation of the flexible internal force after the control of the flexible lander enters the steady state. Combine with the analytical nominal control term established in Step 2 to obtain the analytical nominal-intelligent compensation node height difference control law.

[0112] The specific implementation method of Step 3 is as follows:

[0113] During the simulated landing process of the flexible lander, perform state quantity sampling to generate the data set required for training the neural network. Apply periodic random thrusts in the height direction to the three thrust nodes of the flexible lander. The thrust amplitude of each node is limited, and the value range is -10N ≤ u zi ≤ 10N. The action period of each group of random thrusts is 5s, and then switch to the next group of random thrusts until the pitch angle or roll angle of the flexible lander reaches 30°, or the flight time reaches 50s. During the landing process, sample the node position quantity, node attitude quantity, and internal force received at the node of the flexible lander every 0.05s and record and save the data. Repeat the above landing sampling process 10,000 times to generate the mapping data set of the node state quantity and the node internal force.

[0114] During the landing process of the flexible lander, the internal force received by each node is related to the height difference between nodes, the relative distance between nodes, and the attitude of the node itself. Use the node state quantity and the node internal force difference in the mapping data set as the input quantity and the output quantity, and based on neural network learning, establish the node internal force intelligent mapping function. The expression of the mapping function is:

[0115]

[0116] where f in(·) is the intelligent mapping function, and the node distance deformation variable δL = [L 12 -L0, L 13 -L0, L 23 -L0] T , L 12 is the distance between node 1 and node 2, L 13 is the distance between node 1 and node 3, L 23 is the distance between node 2 and node 3. The node angular acceleration variable is the component of the attitude angular acceleration of node i on the x L axis and y L axis of the landing point coordinate system.

[0117] During the attitude tilting maneuver of the flexible lander, the relative distance between nodes and the node attitude change rapidly, and the nodes are subjected to strong flexible internal forces. At this time, internal force compensation is considered in the control design to make the node height difference control process of the flexible detector have better transient performance. When the flexible lander tends to be stable, the whole lander shows strong rigidity, and the flexible internal force will be over-compensated, resulting in a large steady-state error. Therefore, a weighted switching coefficient of the internal force intelligent compensation term is designed to make the compensation of the flexible internal force more reasonable during the attitude control process.

[0118] For the control process of the attitude tilting maneuver of the flexible lander, the expression of the weighted switching coefficient λ is

[0119]

[0120] where e max = 0.01 m represents the criterion value for the height difference control to enter the steady state. After the attitude maneuver of the flexible lander enters the steady state, the weighted coefficient value switches to zero, and the flexible internal force between nodes is no longer compensated.

[0121] Taking the node thrust difference as the control quantity, the analytical nominal-intelligent compensation node height difference control law is finally written as

[0122]

[0123] Figure 3 and Figure 4 give the step response curves of the analytical nominal-intelligent compensation node height difference control law and the analytical nominal control law without considering the intelligent compensation term when the flexible lander tracks a step maneuver with a pitch angle of 20°. Figure 3 is the node height difference curve, Figure 4This is the pitch angle curve of the flexible lander. It can be seen that the node height difference control law of analytical nominal-intelligent compensation well realizes the tracking of the node height difference and the overall attitude angle. Comparing the dynamic performance indexes, for the tracking of the node height difference, the overshoot of the analytical nominal control is 14.23%, the peak time is 11.2 s, and the steady-state error is 0.015%. While for the intelligent control considering flexible internal force compensation, the overshoot is 9.14%, the peak time is 10.15 s, and the steady-state error is 0.016%. For the tracking of the pitch angle, the overshoot of the analytical nominal control is 14.99%, the peak time is 11.2 s, and the steady-state error is 0.016%. While for the intelligent control considering flexible internal force compensation, the overshoot is 9.97%, the peak time is 10.25 s, and the steady-state error is 0.018%. In comparison, the node height difference control law of analytical nominal-intelligent compensation significantly reduces the overshoot and response time at the cost of a very small steady-state error. The attitude maneuver control of the flexible lander has good transient and steady-state performances.

[0124] Step 4: Design the node thrust distribution law to distribute the node thrust difference control quantity in the node height difference control law of analytical nominal-intelligent compensation established in Step 3 to each node thruster, so that each node thruster with limited thrust amplitude can make full use of its own control ability to realize the intelligent cooperative control of the overall attitude of the flexible lander.

[0125] The specific implementation method of Step 4 is as follows:

[0126] Design the node thrust distribution law to distribute the node thrust difference control quantity to the thrusters of each node. To make full use of the control ability of each node thruster, the maximum value of the node thrust and the minimum value of the node thrust should be opposite to each other, that is, u i =-u k when u i ≤u j ≤u k , i, j, k = 1, 2, 3. The expression of the node thrust distribution law is

[0127]

[0128] Figure 5 shows the three-node thrust curves distributed by the node thrust distribution law when the flexible lander tracks a step maneuver with a pitch angle of 20° according to the node height difference control law of analytical nominal-intelligent compensation. It can be seen that under the condition of limited thrust amplitude, no control saturation occurs. The above distribution law reasonably distributes the thrust difference control quantity, enabling the three-node thrusters to make full use of their own control ability and realizing the cooperative control of the overall attitude of the flexible lander.

[0129] The specific description above further elaborates on the purpose, technical solution, and beneficial effects of the invention. It should be understood that the above description is only a specific embodiment of the present invention and is not used to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A flexible lander intelligent attachment control method, characterized in that: It includes the following steps: Step 1: Describe the overall attitude of the flexible lander by the height difference between the thrust nodes, deduce the conversion relationship between the pitch and roll attitudes of the flexible lander and the node height difference, and transform the attitude control problem of the flexible lander into the control problem of the node height difference. Step 2: Establish a node-oriented dynamic model of the flexible lander, and design an analytical nominal-intelligent compensation node height difference control law with the node thrust difference as the control variable. Based on the PD controller, realize the feedback control of the node height difference and the change rate of the node height difference, and determine the analytical nominal control term in the node height difference control law. Step 3: Based on neural network learning, establish an intelligent mapping function between the node state variables and the node flexible internal force difference as the intelligent compensation term in the node height difference control law; through the weighted switching coefficient of the intelligent compensation term, avoid over-compensation of the flexible internal force after the control of the flexible lander enters the steady state; combine with the analytical nominal control term established in Step 2 to obtain the analytical nominal-intelligent compensation node height difference control law. Step 4: Design a node thrust distribution law, and distribute the node thrust difference control variable in the analytical nominal-intelligent compensation node height difference control law established in Step 3 to each node thruster, so that each node thruster with limited thrust amplitude can make full use of its own control ability to realize the intelligent cooperative control of the overall attitude of the flexible lander.

2. The intelligent attachment control method of the flexible lander according to claim 1, characterized in that: The specific implementation method of Step 1 is as follows: On the flexible lander, there are mass concentration areas equipped with thrusters and sensors, which are called nodes; in the nominal state without deformation, the three-node flexible lander is in the shape of a standard disk; the three nodes are evenly distributed on the surface of the detector, forming an equilateral triangle; the three nodes are numbered 1, 2, and 3 in counterclockwise order, called Node 1, Node 2, and Node 3. On the premise of retaining the flexible characteristics and the planar configuration characteristics, establish a simplified model of the flexible lander oriented to nodes. Define the landing point coordinate system o L -x L y L z L , with the landing point o L as the origin, the normal direction perpendicular to the surface of the small celestial body at the landing point is the z L axis, the eastward direction on the surface of the small celestial body is the x L axis, and the northward direction on the surface is the y L axis; In the landing point coordinate system, the node position vector r i is r i = [x i , y i , z i T (1)​ where i is the thrust node number (i = 1, 2, 3), x i is the component of the line connecting node i and the landing point on the x L axis, y i is the component of the line connecting node i and the landing point on the y L axis, z i is the component of the line connecting node i and the landing point on the z L axis, i.e., the height of node i on the surface of the small celestial body; The mass of the flexible lander is concentrated at each node, and the center of mass of the flexible lander is the centroid of the node triangle. where r c is the centroid position vector of the flexible lander; For the flexible lander with a circular disk-shaped configuration, only consider the attitude tilt along the radial direction of the circular disk. Taking the line connecting the node centroids as the axis, define the pitch attitude and roll attitude of the flexible lander; taking the vector r from the centroid of the flexible lander to node 1 c1 as the axis, the pitch attitude of the flexible lander is represented by the angle between r c1 and the o L x L y L plane. The expression for the pitch angle θ is The roll attitude of the flexible lander is represented by the vector r from node 3 to node 2 32 and o L x L y L plane, and the expression for the roll angle ψ is During the landing process, the change in the distance between nodes is much smaller than the original length of the node distance. When calculating the pitch and roll attitude angles, the change in the distance between nodes is ignored, and the approximate pitch angle and roll angle expressions are where L0 is the original length of the node distance in the nominal state. Define the height difference variable H between the nodes of the flexible lander. Among them, h 12 is the height difference between Node 1 and Node 2, and h 13 is the height difference between Node 1 and Node 3; The pitch and roll attitudes of the disk-shaped flexible lander are characterized by the height difference between the nodes, and the conversion relationship between the height difference variable and the pitch angle and roll angle is Thus, the overall attitude control problem of the flexible lander is transformed into the control problem of the node height difference.

3. The intelligent attachment control method for a flexible lander according to claim 1 or 2, characterized in that: The specific implementation method of Step 2 is as follows: In the landing point coordinate system, the landing dynamics model of the node-oriented flexible lander is where, v i = [v xi , v yi , v zi T is the velocity of node i, u i = [u xi , u yi , u zi T is the thrust of node i, m is the mass of each node, f i = [f xi , f yi , f zi T is the internal flexible force acting on node i, a ai = [a axi , a ayi , a azi T is the acceleration of node i caused by the small celestial body environmental force;​​​​ Since the size of the flexible lander is much smaller than the size of the small celestial body, the external forces of the small celestial body environment received at each node are equal; from the landing dynamics, it is known that the acceleration difference between the node height directions is only related to the thrust difference and the flexible internal force difference in the node height direction, that is Among them, U 12 is the difference in the thrusts received by Node 1 and Node 2 in the height direction, and U 13 is the difference in the thrusts received by Node 1 and Node 3 in the height direction, and f 12 is the difference in the flexible internal forces received by Node 1 and Node 2 in the height direction, and f 13 is the difference in the flexible internal forces received by Node 1 and Node 3 in the height direction, that is From the dynamic relationship between the acceleration difference between the node height directions and the node thrust difference and the flexible internal force difference, design an analytical nominal-flexible compensation node height difference control law with the node thrust difference as the control variable, and the expression is Among them, is the analytical nominal control term of the node height difference, is the compensation term for f 12 , is the compensation term for f 13 ; Based on the PD controller, design the feedback control of the node height difference and the change rate of the node height difference, and establish the analytical nominal control term in the node height difference control law where k p > 0 is the gain of the height difference feedback term, k d > 0 is the gain of the height difference change rate feedback term, H e is the desired height difference, is the desired height difference change rate.

4. The intelligent attachment control method of the flexible lander according to claim 3, wherein: The specific implementation method of Step 3 is During the simulated landing process of the flexible lander, sample the state variables to generate the dataset required for training the neural network; apply periodic random thrusts in the height direction to the three thrust nodes of the flexible lander The thrust amplitude of each node is limited, and the value range is u zi ∈[-u max , u max , where u max is the upper limit value of the thrust constraint; the action period of each group of random thrusts is ΔT, and then the next group of random thrusts is switched until the pitch angle or roll angle of the flexible lander reaches the upper limit value θ max , ψ max , or the flight time reaches the upper limit value T max ; during the landing process, the node position quantity, node attitude quantity and internal force received at the node of the flexible lander are sampled every interval time δt and the data is recorded and saved; the above landing sampling process is repeated N times to generate a mapping data set of the node state quantity and the node internal force; During the landing process of the flexible lander, the internal forces received by each node are related to the height difference between nodes, the relative distance between nodes, and the attitude of the node itself; use the node state variables and the node internal force difference in the mapped dataset as the input and output variables, and based on neural network learning, establish the intelligent mapping function of the node internal force. The expression of the mapping function is Among them, f in (·) is the intelligent mapping function, and the node distance deformation variable δL = [L 12 -L0, L 13 -L0, L 23 -L0] T , L 12 is the distance between node 1 and node 2, L 13 is the distance between node 1 and node 3, L 23 is the distance between node 2 and node 3; the node angular acceleration variable is the component of the attitude angular acceleration of node i along the x L axis and y L axis at the landing point coordinate system; When the flexible lander performs an attitude tilting maneuver, the relative distance between nodes and the node attitude change rapidly, and the node is subjected to strong flexible internal forces. At this time, considering internal force compensation in the control design can make the node height difference control process of the flexible detector have better transient performance; when the attitude of the flexible lander tends to be stable, the flexible lander as a whole exhibits strong rigidity, and the flexible internal force will be overcompensated, resulting in a large steady-state error; therefore, through the weighted switching coefficient of the intelligent internal force compensation term, the compensation of the flexible internal force in the attitude control process is made more reasonable For the control process of the attitude tilting maneuver of the flexible lander, the expression of the weighted switching coefficient λ is Among them, e max represents the criterion value for the height difference control to enter the steady state; after the attitude maneuver of the flexible lander enters the steady state, the weighted coefficient value switches to zero, and the flexible internal force between nodes is no longer compensated; Taking the node thrust difference as the control variable, the node height difference control law of analytical nominal-intelligent compensation is finally written as 5. The intelligent attachment control method of the flexible lander according to claim 4, wherein: The specific implementation method of Step 4 is Design the node thrust distribution law to distribute the control quantity of the node thrust difference to the thrusters of each node; to make full use of the control capabilities of the thrusters of each node, the maximum node thrust value and the minimum node thrust value should be opposite to each other, that is, u i =-u k when u i ≤u j ≤u k , i, j, k = 1, 2, 3; the expression of the node thrust distribution law is According to formula (16), distribute the node thrust difference control variable to each node thruster, so that each node thruster with limited thrust amplitude can make full use of its own control ability to achieve the intelligent cooperative control of the overall attitude of the flexible lander

Images